Semi-Supervised Classification with Graph Convolutional Networks

Benchmark Model Rank Results
graph-classification-on-ddGCN#29Accuracy: 78.151±3.465
graph-classification-on-enzymesGCN#7Accuracy: 73.466±4.372
graph-classification-on-imdb-bGCN#8Accuracy: 79.500±3.109
graph-classification-on-nci1GCN#19Accuracy: 84.185±0.644
graph-classification-on-nci109GCN#11Accuracy: 83.140±1.248
graph-classification-on-proteinsGCN#54Accuracy: 75.536±1.622
graph-property-prediction-on-ogbg-code2GCN+virtual node#7Test F1 score: 0.1595 ± 0.0018Ext. data: No
graph-property-prediction-on-ogbg-code2GCN#14Test F1 score: 0.1507 ± 0.0018Ext. data: No
graph-property-prediction-on-ogbg-molhivGCN#32Test ROC-AUC: 0.7606 ± 0.0097Ext. data: No
graph-property-prediction-on-ogbg-molhivGCN+virtual node#33Test ROC-AUC: 0.7599 ± 0.0119Ext. data: No
graph-property-prediction-on-ogbg-molhivGCN (in Julia)#35Test ROC-AUC: 0.7549 ± 0.0163Ext. data: No
graph-property-prediction-on-ogbg-molpcbaGCN+virtual node#24Test AP: 0.2424 ± 0.0034Ext. data: NoValidation AP: 0.2495 ± 0.0042
graph-property-prediction-on-ogbg-molpcbaGCN#29Test AP: 0.2020 ± 0.0024Ext. data: NoValidation AP: 0.2059 ± 0.0033
graph-property-prediction-on-ogbg-ppaGCN+virtual node#13Ext. data: NoTest Accuracy: 0.6857 ± 0.0061
graph-property-prediction-on-ogbg-ppaGCN#14Ext. data: NoTest Accuracy: 0.6839 ± 0.0084
graph-regression-on-esr2GCN#7R2: 0.642±0.000RMSE: 0.528±0.642
graph-regression-on-f2GCN#7R2: 0.878±0.000RMSE: 0.355±0.878
graph-regression-on-kitGCN#7R2: 0.814±0.000RMSE: 0.469±0.814
graph-regression-on-lipophilicityGCN#7RMSE: 0.565±0.011R2: 0.800±0.008
graph-regression-on-parp1GCN#7R2: 0.912±0.000RMSE: 0.372±0.912
graph-regression-on-pcqm4mv2-lscGCN#19Validation MAE: 0.1379Test MAE: 0.1398
graph-regression-on-pgrGCN#8R2: 0.658±0.000RMSE: 0.565±0.658
graph-regression-on-zinc-fullGCN#19Test MAE: 0.152±0.023
heterogeneous-node-classification-on-acmGCN#5Macro-F1: 92.17Micro-F1: 92.12
heterogeneous-node-classification-on-dblp-2GCN#8Macro-F1: 90.84Micro-F1: 91.47
heterogeneous-node-classification-on-freebaseGCN#7Macro-F1: 27.84Micro-F1: 60.23
heterogeneous-node-classification-on-imdbGCN#8Macro-F1: 57.88Micro-F1: 64.82
link-property-prediction-on-ogbl-citation2Full-batch GCN#11Ext. data: NoTest MRR: 0.8474 ± 0.0021Validation MRR: 0.8479 ± 0.0023
link-property-prediction-on-ogbl-collabGCN (val as input)#16Test Hits@50: 0.4714 ± 0.0145Ext. data: No
link-property-prediction-on-ogbl-collabGCN#18Test Hits@50: 0.4475 ± 0.0107Ext. data: No
link-property-prediction-on-ogbl-ddiGCN+JKNet#13Ext. data: NoTest Hits@20: 0.6056 ± 0.0869
link-property-prediction-on-ogbl-ddiGCN#15Ext. data: NoTest Hits@20: 0.3707 ± 0.0507
link-property-prediction-on-ogbl-ppaGCN#15Ext. data: NoTest Hits@100: 0.1867 ± 0.0132
molecular-property-prediction-on-esolGCN#4RMSE: 0.520±0.024R2: 0.936±0.006
molecular-property-prediction-on-freesolvGCN#6RMSE: 0.815±0.086R2: 0.957±0.009
node-classification-on-brazil-air-trafficGCN_cheby (Kipf and Welling, 2017)#3Accuracy: 0.516
node-classification-on-chameleon-60-20-20GCN#181:1 Accuracy: 64.18 ± 2.62
node-classification-on-citeseerGCN#48Accuracy: 70.3
node-classification-on-citeseer-60-20-20GCN#151:1 Accuracy: 81.39 ± 1.23
node-classification-on-coraGCN#49Accuracy: 81.5%
node-classification-on-cora-60-20-20-randomGCN#191:1 Accuracy: 87.78 ± 0.96
node-classification-on-cornell-60-20-20GCN#251:1 Accuracy: 82.46 ± 3.11
node-classification-on-europe-air-trafficGCN_cheby (Kipf and Welling, 2017)#2Accuracy: 46.0
node-classification-on-europe-air-trafficGCN (Kipf and Welling, 2017)#6Accuracy: 37.1
node-classification-on-facebookGCN_cheby (Kipf and Welling, 2017)#5Accuracy: 64.6
node-classification-on-facebookGCN (Kipf and Welling, 2017)#7Accuracy: 57.5
node-classification-on-film-60-20-20-randomGCN#291:1 Accuracy: 35.51 ± 0.99
node-classification-on-flickrGCN (Kipf and Welling, 2017)#6Accuracy: 0.546
node-classification-on-flickrGCN_cheby (Kipf and Welling, 2017)#7Accuracy: 0.479
node-classification-on-geniusGCN#14Accuracy: 87.42 ± 0.37
node-classification-on-non-homophilicGCN#251:1 Accuracy: 82.46 ± 3.11
node-classification-on-non-homophilic-1GCN#231:1 Accuracy: 75.5 ± 2.92
node-classification-on-non-homophilic-13GCN#111:1 Accuracy: 82.47 ± 0.27
node-classification-on-non-homophilic-14GCN#161:1 Accuracy: 87.42 ± 0.37
node-classification-on-non-homophilic-15GCN#201:1 Accuracy: 62.18 ± 0.26
node-classification-on-non-homophilic-2GCN#221:1 Accuracy: 83.11 ± 3.2
node-classification-on-non-homophilic-4GCN#171:1 Accuracy: 64.18 ± 2.62
node-classification-on-non-homophilic-6GCN#211:1 Accuracy: 62.23±0.53
node-classification-on-penn94GCN#15Accuracy: 82.47 ± 0.27
node-classification-on-pubmedGCN#42Accuracy: 79.0
node-classification-on-pubmed-60-20-20-randomGCN#231:1 Accuracy: 88.9 ± 0.32
node-classification-on-squirrel-60-20-20GCN#221:1 Accuracy: 44.76 ± 1.39
node-classification-on-texas-60-20-20-randomGCN#241:1 Accuracy: 83.11 ± 3.2
node-classification-on-wisconsin-60-20-20GCN#261:1 Accuracy: 75.5 ± 2.92
node-property-prediction-on-ogbn-arxivGCN+residual+6 layers#45Test Accuracy: 0.7286 ± 0.0016Ext. data: No
node-property-prediction-on-ogbn-arxivGCN+residual+node2vec#46Test Accuracy: 0.7278 ± 0.0013Ext. data: No
node-property-prediction-on-ogbn-arxivGCN_res + 8 layers#50Test Accuracy: 0.7262 ± 0.0037Ext. data: No
node-property-prediction-on-ogbn-arxivGCN#67Test Accuracy: 0.7174 ± 0.0029Ext. data: No
node-property-prediction-on-ogbn-productsFull-batch GCN#52Test Accuracy: 0.7564 ± 0.0021Ext. data: No
node-property-prediction-on-ogbn-proteinsGCN#20Ext. data: NoTest ROC-AUC: 0.7251 ± 0.0035